기본 콘텐츠로 건너뛰기

Mathematical Tools for Computer Vision

University of Science and Technology
Electronics and Telecommunications Research Institute (ETRI)
Vision System Research Team, Jae-Young LEE
March 1, 2016 ~ June 24, 2016


   - Vector & Scalar & Vector Equality, Addition, and Subtraction
   - Euclidean Vector
   - Trigonometry Review
   - Polar Representation
   - Dot product & Cross product
   - Homework #1
   - Homework #2



2. Geometry: 공간 도형
   - 도형
   - 직선의 방정식
   - 평면의 방정식
   - 부등식의 영역
   - 도형의 방정식과 함수
   - Homework #3



3. 행렬 연산
   - Introduction
   - Matrices and matrix algebra
   - Matrices and systems of linear equations



4. 벡터공간, 선형시스템
   - Vector Spaces
   - Basis and Dimension
   - Rank of a Matrix and Systems of Linear Equations
   - 기타 선형대수학에서 알아두어야 할 것들
   - Homework #4
   - Homework #5



   - Eigenvalue and Eigenvector
   - Eigendecomposition
   - PCA (Principal Component Analysis)
   - SVD (Singular Value Decomposition)



6. 파라미터 추정과 최소 자승법
   - 파라미터 추정 기초
   - 최소 자승법 기초
   - Homework #8 (최소자승법을 이용한 영상 이진화)



7. Robust 파라미터 추정 기법
   - M-estimator (Weighted LS)
   - RANSAC
   - Homework #9 (LS, M-estimtor, RANSAC을 이용한 불균일 영역 검출)



8. 최적화 기법
   - 일차미분(Gradient descent)을 이용한 최적화
   - 이차미분(Newton)을 이용한 최적화 (2차 Taylor 근사)
   - 최적화 보완 기법 (Trust Region, Line Search, Saddle-free Newton)
   - 다변수함수에서의 최적화
   - LS 문제에 특화된 기법 (Gauss-Newton, Levenberg-Marquardt)
   - Homework #10 (원 근사 or 최적화, Gradient descent와 Gauss-newton 방법 사용)



9. 기계 학습 기법
   - 기계학습 기초
   - Gaussian Mixture Model 등 여러가지 알고리즘





Review
After taking this course, I can get knowledge of linear algebra such as vectors, matrices, eigen-decomposition, parameter estimation, Principal Component Analysis, Singular Value Decomposition, optimization techniques and so on and the computer vision. Actually, every class, we learn some linear algebra and we apply it to computer vision domain in order to solve the problem. It's very interesting because it is not only mathematics, but it's real world techniques of computer vision problems by using OpenCV or MATLAB. In this process, I could easily understand the concepts and have even intuition for difficult concepts.

One of the important knowledge is a way of solving linear or non-linear algebra system. Since in the real world, there are many problems of linear algebra system. So, given the knowledge, I could represent the problem to matrices (linear algebra representation) and know how to solve it even the non-linear case.

Taking this opportunity, the interest of linear algebra is increased more and more. I want to deep understand of the concepts. Linear algebra field is very interesting.  


Reference
[1] 다크프로그래머, http://darkpgmr.tistory.com/





댓글

이 블로그의 인기 게시물

Pattern Discovery in Data Mining

Coursera Illinois at Urbana-Champaign by Jiawei Han 2015.03.19 CONTENT 1. A brief Introduction to Data Mining 2. Pattern Discovery : Basic Concepts 3. Efficient Pattern Mining Methods 4. Pattern Evaluation 5. Mining Diverse Patterns 6. Constraint-Based Pattern Mining 7. Sequential Pattern Mining 8. Graph Pattern Mining 9. Pattern-Based Classification 10. Exploring Pattern Mining Applications Lecture 1 : A brief Introduction to Data Mining - We'are drowning in data but starving for knowledge ( a lot of data are unstructured ) - Data mining : a misnomer ! -> Knowledge mining from data - Extraction of interesting patterns (non-trivial, implicit, previously unknown and potentially useful) or knowledge from massive data. - Data mining is a interdisciplinary field (machine learning, pattern recognition, statistics, databases, big data, business intelligence..) Knowledge Discovery (KDD) Process Methodology View: Confluence of Multiple Disciplines Lecture 2 : Pattern Discovery : Ba...

Vector Space Model

Motivation When you want to find some information by using Search Engines, you have to make a query used for search. Unfortunately, since you don't know exactly what it means, your query will be ambiguous and not accurate. Therefore, Search Engines give you the information in a ranked list rather than the right position. Intuition In order to make a ranked list, you need to calculate the similarity between the query and documents based on terms or words. One of the calculation of similarity is dot product on a vector space. In the vector space, there are many documents with respect to word dimensions The first to rank is d2, because to see with eyes it's the most similarity with the query. Problem How do we plot those vectors wonderfully and nicely and very very fairly ? - How do we define the dimension ? - How do we place a document vector ? - How do we place a query vector ? - How do we match a similarity ? Consideration 1. The frequency of each word of Query. First, Score in...

Operating System

Operating System Three Phases of OS History Phase 1       Summary :  H/W 비싸고, 인건비는 저렴        Goal : 어떻게 하면 값비싼 CPU를 효율적으로 사용할 수 있을까 ? Phase 1-1 : Operator as OS 하드웨어(애니악 - 진공관 CPU,마그네틱 코어 Memory)가 매우 비싸고, 그에 반해 인건비가 매우 저렴하기 때문에 CPU Utilization을 최대화 시키는게 목표였다. 초창기 컴퓨터 당시 Operator 역할은 사람이었다. 카드 덱 수령, 카드 덱 시스템에 로딩, 수행결과를 프린트, 결과물을 사용자에 전달 등등 Phase 1-2 : Simple batch monitor Human Operator에 의해 job-to-job 전환 속도가 매우 느린점으로 인해 CPU적 관점으로 컴퓨터 시스템의 비효율성을 극복(Utilization 최대화)하기 위해서 최초의 Operating System Software가 등장했다. I/O machine이 추가 되었다. Phase 1-3 : Batch monitor ( OS + I/O device controller ) 하지만, I/O를 추가한 후, I/O를 하고 있을 때 CPU가 idle하게 되었고, 또 다시 CPU Utilization 문제가 대두되었다. 이와 같은 문제를 해결하기 위해 I/O Channel, I/O device controller가 등장했다. I/O Channel은 CPU를 대신해서 I/O device의 operation을 관장해준다. 그 대신 I/O operation의 시작과 끝만 CPU에게 알려준다. ( Asynchronous I/O 방법으로써 Interrupt mechanism 등장 ) 이러한 mechanism 으로 CPU와 I/O가 서로 overlap되며 사용가능해졌다. Phase 1-4 : Multi...